Timezone: »

Regret Analysis for Continuous Dueling Bandit
Wataru Kumagai

Tue Dec 05 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #5
The dueling bandit is a learning framework where the feedback information in the learning process is restricted to noisy comparison between a pair of actions. In this paper, we address a dueling bandit problem based on a cost function over a continuous space. We propose a stochastic mirror descent algorithm and show that the algorithm achieves an $O(¥sqrt{T¥log T})$-regret bound under strong convexity and smoothness assumptions for the cost function. Then, we clarify the equivalence between regret minimization in dueling bandit and convex optimization for the cost function. Moreover, considering a lower bound in convex optimization, it is turned out that our algorithm achieves the optimal convergence rate in convex optimization and the optimal regret in dueling bandit except for a logarithmic factor.

Author Information

Wataru Kumagai (RIKEN)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors